Abstract
Abstract Prior knowledge of the performance of a tropical cyclone intensity forecast holds the potential to increase the value of forecasts for end users. The values of certain dynamical parameters, such as storm speed, latitude, current intensity, potential intensity, wind shear magnitude, and direction of the shear vector, are shown to be related to the error of an individual model forecast. The varying success of each model in the different environmental conditions represents a source of additional information on the reliability of an individual forecast beyond average forecast error. Three hurricane intensity models that were operational for the duration of the five hurricane seasons between 2006 and 2010, as well as the National Hurricane Center official forecast (OFCL), are evaluated for 24-, 48-, and 72-h forecasts in the Atlantic Ocean. The performance of each model is assessed by computing the mean absolute error, bias, and percent skill relative to a benchmark model. The synoptic variables are binned into physically meaningful ranges and then tested individually and in combinations to capture the different regimes that are conducive to forecasts with higher or lower error. The results address conventional wisdom about which environmental conditions lead to better forecasts of hurricane intensity and highlight the different strengths of each model. The statistical significance established between the different bins in each model as well as the corresponding bins for other models indicates there is the potential for error predictions to accompany tropical cyclone intensity forecasts.
